CV
Download the PDF version with the button on the right.
Contact Information
| Name | Mihajlo Stevanović |
| Professional Title | MSc MVA student, ENS Paris-Saclay · Robot learning |
| mihastevanovic04@gmail.com |
Professional Summary
Robot learning (imitation and reinforcement learning), world models, contact-rich and dynamic manipulation, trajectory optimization and model-based control, sim-to-real transfer.
Experience
-
2026 - present Belgrade, Serbia
Research project: robotic table tennis, ball interception (in progress)
ETF Robotics Laboratory, University of Belgrade
- Gazebo setup on franka_ros (ROS 1 Noetic); MoveIt interceptor within 1–2 cm of the predicted contact point but too slow, so developing a nonlinear MPC (CasADi/IPOPT, acados).
-
2026 - 2026 Zrenjanin, Serbia
Selected participant
PSIML, Practical Seminar in Machine Learning
- Action-conditioned video world model: LoRA fine-tuning of Wan2.1 on BAIR robot pushing; 84.8% direction accuracy on 256 held-out scenes.
-
2025 - 2026 Belgrade, Serbia
Robotics Research Intern
ETF Robotics Laboratory, University of Belgrade
- IL/RL policies for cable insertion in Isaac Lab for the AI for Industry Challenge (Intrinsic & Open Robotics).
- OpenUSD scenes and a containerized ROS 2 / Gazebo sim-to-real environment.
- BSc thesis: 30,290 demos, ACT vs. Diffusion Policy data-scaling study (lateral error is the main obstacle; 3.5× lower with more data: 41.2 → 11.8 mm).
-
2025 - 2026 Belgrade, Serbia
Teaching Assistant (lab sessions, 4 undergraduate courses)
School of Electrical Engineering, University of Belgrade
Education
-
2026 - 2027 Gif-sur-Yvette, France
MSc (M2)
École Normale Supérieure (ENS) Paris-Saclay
Mathématiques, Vision, Apprentissage (MVA)
- French Government Scholarship (BGF)
- Fall courses: Robotics · Convex Optimization · Probabilistic Graphical Models & Deep Generative Models · Optimal Transport for ML · Foundations of LLMs
-
2022 - 2026 Belgrade, Serbia
BSc
University of Belgrade, School of Electrical Engineering
Electrical Engineering and Computing, Signals and Systems
- Thesis: Limits of Imitation Learning for High-Precision Manipulation (advisor: Prof. Kosta Jovanović)
- Selected courses: Robotics & Automation · Autonomous Mobile Robots · Control Systems 1 & 2 · System Modeling & Identification · Stochastic Systems & Estimation · Neural Networks
Publications
-
2026 Creating a virtual environment to bridge the gap between simulation and reality in robot assembly tasks
Proc. LXX ETRAN Conference, pp. 173–177 (in Serbian)
Awards
-
2026 French Government Scholarship (BGF)
Government of France
For the M2 MVA at ENS Paris-Saclay.
-
2026 3rd Place, Play With IT Agri-Bot Hackathon (team lead)
Embassy of Italy, Belgrade
Skills
Robotics: NVIDIA Isaac Sim/Lab, OpenUSD, ROS 1 (Noetic), ROS 2 (Jazzy, Kilted), Gazebo, ABB RobotStudio
Machine learning: PyTorch, ACT, Diffusion Policy, reinforcement learning, LoRA fine-tuning, NumPy, SciPy
Programming: Python, C, C++, MATLAB, Bash, Git, Docker, Linux
Languages
Serbian : Native
English : C1 (TOEFL iBT 90)
German : B1
French : A2